<p>This paper proposes an advanced modeling approach to identify the compressive strength of High-Performance Concrete (HPC), the interpretation of features, and the sustainability involved in its creation process regarding values of CO₂ emissions and energy intake. In this respect, to reach a finer understanding of the relationships between the contributions of each covariate concerning the target variable, Partial Dependence Plots (PDP) were applied within this research exercise. The present framework is designed to incorporate interpretability, other than traditional Machine Learning (ML) approaches that focus solely on model predictive accuracy and estimate the environmental and energy-related impacts of the critical input variables such as cement content, which was identified as the most energy-intensive and CO₂ In this regard, for further optimization of feature selection and to improve the performance of the K-Nearest Neighbor (KNN) model, some advanced metaheuristic algorithms are implemented, including Sand Cat Swarm Optimization (SCSO), Gradient-Based Optimizer (GBO) and Brown Bear Optimization Algorithm (BBOA). The proposed framework demonstrated robust predictive capabilities with the KNN model, including an R² of 0.962 at training time, 0.948 at validation time, and 0.945 at test time. By coupling interpretability with advanced optimization, the proposed framework establishes a strong dual focus. It enhances model accuracy while simultaneously addressing key environmental and energy efficiency challenges associated with HPC design and production.</p>